From efc6875f6cb42735bb9cd277d7137c1333d5053e Mon Sep 17 00:00:00 2001
From: phoenixAja <plogan@chanzuckerberg>
Date: Mon, 17 Jun 2024 12:58:42 -0700
Subject: [PATCH] subsample with seqtk instead of using head

---
 workflows/long-read-mngs/run.wdl | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/workflows/long-read-mngs/run.wdl b/workflows/long-read-mngs/run.wdl
index 410e9e394..e1ffb2a16 100644
--- a/workflows/long-read-mngs/run.wdl
+++ b/workflows/long-read-mngs/run.wdl
@@ -268,7 +268,9 @@ task RunSubsampling {
 
     command <<<
         set -euxo pipefail
-        head -"~{subsample_depth}" "~{input_fastq}" > sample.subsampled.fastq
+
+        # set seed to 42 for reproducibility
+        seqtk sample -s42 "~{input_fastq}" "~{subsample_depth}" > sample.subsampled.fastq
 
         # We should always have reads after subsampling, but adding for consistency with other steps
         filter_count sample.subsampled.fastq subsampled "No reads remaining after subsampling"